mlsanityData to 7 Oct 2026

Language Mixing in Reasoning Language Models: Patterns, Impact, and Internal Causes

arXiv 2505.14815 ↗First posted 20 May 2025cs.CLcited by 20

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Reasoning language models (RLMs) excel at complex tasks by leveraging a chain-of-thought process to generate structured intermediate steps. However, language mixing, i.e., reasoning steps containing tokens from languages other than the prompt, has been observed in their outputs and shown to affect performance, though its impact remains debated. We present the first systematic study of language mixing in RLMs, examining its patterns, impact, and internal causes across 15 languages, 7 task difficulty levels, and 18 subject areas, and show how all three factors influence language mixing. Moreover, we demonstrate that the choice of reasoning language significantly affects performance: forcing models to reason in Latin or Han scripts via constrained decoding notably improves accuracy. Finally, we show that the script composition of reasoning traces closely aligns with that of the model's internal representations, indicating that language mixing reflects latent processing preferences in RLMs. Our findings provide actionable insights for optimizing multilingual reasoning and open new directions for controlling reasoning languages to build more interpretable and adaptable RLMs.

Key facts. Cited by 20 papers in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 13 of its 34 references that are in the map. First posted on arXiv 20 May 2025.

Builds on

13 of 34 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025Crosslingual Reasoning through Test-Time Scaling8 May
  2. 2024On Memorization of Large Language Models in Logical Reasoning30 Oct

Other references · 11

  1. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  2. 2016Bag of Tricks for Efficient Text Classificationcited by 869
  3. 2024Do Llamas Work in English? On the Latent Language of Multilingual Transformerscited by 275
  4. 2025The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Taskscited by 150
  5. 2020GLUECoS : An Evaluation Benchmark for Code-Switched NLPcited by 101
  6. 2024Understanding and Mitigating Language Confusion in LLMscited by 75
  7. 2023Multilingual Large Language Models Are Not (Yet) Code-Switcherscited by 65
  8. 2023A Survey of Code-switching: Linguistic and Social Perspectives for Language Technologiescited by 59
  9. 2023Prompting Multilingual Large Language Models to Generate Code-Mixed Texts: The Case of South East Asian Languagescited by 50
  10. 2025Lost in Multilinguality: Dissecting Cross-lingual Factual Inconsistency in Transformer Language Modelscited by 43
  11. 2021More but Correct: Generating Diversified and Entity-revised Medical Responsecited by 31

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 9; 2026: 11.

20 papers.

Sort:

By area:

  1. 2026Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Modelscited by 22
  2. 2025Think Natively: Unlocking Multilingual Reasoning with Consistency-Enhanced Reinforcement Learningcited by 17
  3. 2025Why Do Multilingual Reasoning Gaps Emerge in Reasoning Language Models?cited by 11
  4. 2025Mechanistic Understanding and Mitigation of Language Confusion in English-Centric Large Language Modelscited by 8
  5. 2025A Comprehensive Evaluation of Multilingual Chain-of-Thought Reasoning: Performance, Consistency, and Faithfulness Across Languagescited by 6
  6. 2025What Language(s) Does Aya-23 Think In? How Multilinguality Affects Internal Language Representationscited by 4
  7. 2025Parallel Scaling Law: Unveiling Reasoning Generalization through A Cross-Linguistic Perspectivecited by 4
  8. 2025Gradual Code-Switching as Inference-Time Cross-Lingual Representational Alignment for LLMscited by 3
  9. 2025SASFT: Sparse Autoencoder-guided Supervised Finetuning to Mitigate Unexpected Code-Switching in LLMscited by 2
  10. 2026OLA: Output Language Alignment in Code-Switched LLM Interactionscited by 2
  11. 2026Think Multilingual, Not Harder: A Data-Efficient Framework for Teaching Reasoning Models to Code-Switchcited by 2
  12. 2026Steering the Language Axis: From Linear Decodability to Causal Controlcited by 1
8 more
  1. 2025Challenging Multilingual LLMs: A New Taxonomy and Benchmark for Unraveling Hallucination in Translationcited by 0
  2. 2026Neither Here Nor There: Cross-Lingual Representation Dynamics of Code-Mixed Text in Multilingual Encoderscited by 0
  3. 2026Identifying Influential N-grams in Confidence Calibration via Regression Analysiscited by 0
  4. 2026HiMed: Incentivizing Hindi Reasoning in Medical LLMscited by 0
  5. 2026Understanding LLM Behavior in Multi-Target Cross-Lingual Summarizationcited by 0
  6. 2026Language-Aware Token Boosting: LLM Language Confusion Reduction Without Tuningcited by 0
  7. 2026When Safety Speaks a Language: A Mechanistic Analysis of Safety-Language Identity Entanglement in LLMscited by 0
  8. 2026When Does Reasoning Help in Machine Translation? A Hierarchical Analysis of LRM Reasoning Tracescited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Language Matters: How Do Multilingual Input and Reasoning Paths Affect Large Reasoning Models?
  2. 2026Rethinking the Multilingual Reasoning Gap with Layer Swap
  3. 2025Beg to Differ: Understanding Reasoning-Answer Misalignment Across Languages
  4. 2025The Impact of Language Mixing on Bilingual LLM Reasoning
  5. 2025When Models Reason in Your Language: Controlling Thinking Language Comes at the Cost of Accuracy
  6. 2024Beyond Accuracy: Evaluating the Reasoning Behavior of Large Language Models -- A Survey
  7. 2026What Makes Good Multilingual Reasoning? Disentangling Traces with Measurable Features
  8. 2025When Less Language is More: Language-Reasoning Disentanglement Makes LLMs Better Multilingual Reasoners